| Summary: | libcairo 1.2.0 much much slower than 1.0.4 on an ssh connection | ||
|---|---|---|---|
| Product: | cairo | Reporter: | Claudio Sacerdoti Coen <sacerdot> |
| Component: | xlib backend | Assignee: | Carl Worth <cworth> |
| Status: | RESOLVED FIXED | QA Contact: | cairo-bugs mailing list <cairo-bugs> |
| Severity: | blocker | ||
| Priority: | high | CC: | baptiste.millemathias |
| Version: | 1.2.0 | ||
| Hardware: | x86 (IA32) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
Seems like the bug reported here: http://lists.freedesktop.org/archives/cairo/2006-July/007434.html Unnecessary Xflush calls are being made with cairo 1.2. Will be fixed in 1.2.2. The fix for the XSync bug has now been pushed out into cairo "1.2.1": http://gitweb.freedesktop.org/?p=cairo;a=commit;h=b7191885c88068dad57d68ced69a752d1162b12c Claudio, Can you please test that that fix does indeed fix the problem you noticed? You could do this by pulling the latest version of cairo from git, or else by downloading the patch from the URL quoted above, and applying it. If we don't hear feedback one way or the other by the time all the other blocker bugs are fixed, then we'll assume that this bug is fixed and we'll release 1.2.2. Thanks, -Carl Dear Carl,
indeed the patch solves the problem. I think you can close this bug.
Cheers,
C.S.C.
Thanks! |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
Upgrading libcairo 1.0.4 to 1.2.0 we noticed that an application of ours, run over an SSH connection with X forwarding enabled, is between 12 and 15 times slower (hence unusable). The application does intensive use of text rendering via gtksourceview and gtkmathview. Investigating the reason, we realized that every application that does intensive use of text rendering (e.g. gvim) is much slower too! (just open with gvim a big file on a slow connection and you will see by yourself). Comparing network traffic using our application, we noticed that: 1. the overall size of the exchanged packets with version 1.2.0 is only slightly bigger than 1.0.4. 2. the overall number of exchanged packets with version 1.2.0 is about 6 times bigger. We suspect 2 to be a possible cause because of the latency of every packet during the roundtrip. Other infos: we are actually running the testing version of Debian. The phenomenon can be reproduced both on PC with hardware acceleration and on PC wihtout it.